Oracle's Ambitious AI Supercluster: 130,000 NVIDIA GPUs and Nuclear-Powered Data Centers

Oracle to use 130,000+ NVIDIA Blackwell AI GPUs supercluster, powered by 3 nuclear reactors

Oracle has announced it's spending over $100 billion on 2000+ new data centers, expanding on the 160 data centers in operation, with NVIDIA getting 40% of that business for AI hardware. Not only that, but not one, not two, but three nuclear reactors could power the new Blackwell AI GPU supercluster. NVIDIA and Oracle will be launching zettascale OCI superclusters with over 100,000 AI GPUs, with new infrastructure to accelerate AI training and deployment of generative AI models. NVIDIA GB200 liquid-cooled bare-metal instances for large-scale AI applications will be introduced, with Oracle to offer NVIDIA HGX H200 Tensor Core GPUs, connecting up to 65,536 AI GPUs for real-time inference. Oracle CEO said: "So we're in the middle of designing a data center that's north of the gigawatt that has -- but we found the location and the power place we look at it, they've already got building permits for 3 nuclear reactors. These are the small modular nuclear reactors to power the data center. This is how crazy it's getting. This is what's going on". The company explains: "So that goes on, and we'll see more and more applications look at that. So I wouldn't -- if your horizon is over the next 5 years, maybe even the next 10 years, I wouldn't worry about, hey, we've now trained all the models we need and all we need to do is inferencing. I think this is an ongoing battle for technical supremacy that will be fought by a handful of companies and maybe one nation state over the next 5 years at least, but probably more like 10. So this business is just growing larger and larger and larger. There's no slowdown or shift coming". Oracle CEO Larry Elison talked about AI market growth, where he said: "I mean these AI models, these frontier models are going to -- the entry price for a real frontier model from someone who wants to compete in that area is about $100 billion. Let me repeat, around $100 billion. That's over the next 4, 5 years for anyone who wants to play in that game. That's a lot of money. And it doesn't get easier. So there are not going to be a lot of those. I mean we -- this is not the place the list who can actually build one of these frontier models". "But in addition to that, there are going to be a lot of very, very specialized models. I can tell you things that I'm personally involved in, which are using computers to look at, biopsies of slides or CAT scans to discover cancer. Also, there are also blood tests were for discovery and cancer. Those tend to be very specialized models. Those tend not necessarily use the foundational the rocks and the ChatGPTs, and the Gemini, they tend to be highly specialized models. Trained on image recognition on certain data, I mean, literally millions of biopsy slides, for example, and not much other training data is helpful".

Oracle CEO says company is spending $100+ billion on 2000+ data centers, NVIDIA gets 40% of it

Oracle is pushing all-in with the data center market, promising to spend $100+ billion over the next 4 years, with NVIDIA to get 40% of that $100B as it's the global leader in AI GPUs. Right now, Oracle has 162 cloud data centers in operation and under construction across the world, explains Oracle chairman and CTO, Larry Ellison. He said: "the largest of these datacenters is 800 megawatts and will contain acres of NVIDIA GPU Clusters for training large scale AI models". He continued, adding: "Oracle could operate up to 2000 data centers in the future, a significant increase from the 162 currently in operation". It's not just this news, but the Oracle CEO added: "So we're in the middle of designing a data center that's north of the gigawatt that has -- but we found the location and the power place we look at it, they've already got building permits for 3 nuclear reactors. These are the small modular nuclear reactors to power the data center. This is how crazy it's getting. This is what's going on". Yes, you read that right -- building permits for not one, but 3 nuclear reactors... they're small modular nuclear reactors, making me wonder if we'll see Rolls Royce's new Micro-Reactors, new zero-emission power using advanced nuclear technology. You can read more about those in the link below: Oracle CEO Larry Elison talked about AI market growth, where he said: "I mean these AI models, these frontier models are going to -- the entry price for a real frontier model from someone who wants to compete in that area is about $100 billion. Let me repeat, around $100 billion. That's over the next 4, 5 years for anyone who wants to play in that game. That's a lot of money. And it doesn't get easier. So there are not going to be a lot of those. I mean we -- this is not the place the list who can actually build one of these frontier models". "But in addition to that, there are going to be a lot of very, very specialized models. I can tell you things that I'm personally involved in, which are using computers to look at, biopsies of slides or CAT scans to discover cancer. Also, there are also blood tests were for discovery and cancer. Those tend to be very specialized models. Those tend not necessarily use the foundational the rocks and the ChatGPTs, and the Gemini, they tend to be highly specialized models. Trained on image recognition on certain data, I mean, literally millions of biopsy slides, for example, and not much other training data is helpful".

Oracle will use three small nuclear reactors to power new 1-gigawatt AI data center

Oracle has secured the permits to build three small modular reactors (SMRs) to power its AI data center. During its quarterly earnings call, the company said (via The Register) that it plans to use those tiny nuclear plants for a planned AI data center with at least one-gigawatt capacity. SMRs are miniaturized reactors similarly sized to those used on naval vessels like submarines and aircraft carriers. However, since they do not have to be built inside the cramped space of a warship, SMRs do not have to be customized to the needs of a particular vessel. This means Oracle could find a supplier to mass produce it for them at a lower cost than the Navy. Furthermore, an SMR's modular design means that it should, in theory, be cheaper to operate, especially as it no longer has the massive infrastructure often associated with traditional nuclear power plants. However, there's still the question of whether the surrounding area where Oracle plans to deploy its SMRs would allow it. After all, nuclear power plants have this stigma about them, especially after the highly publicized meltdowns at Chernobyl and Fukushima, Japan. Besides, securing the permits to build SMRs is likely different from getting the green light to operate them. Aside from safety concerns, Oracle would also need to contend with security issues, especially as nuclear material is hazardous, especially if it falls into the wrong hands. Nevertheless, Oracle is likely keen on getting this project up and running. "Oracle has 162 cloud data centers, live and under construction worldwide. The largest of these data centers is 800 megawatts, and it will contain acres of Nvidia GPU clusters able to train the world's largest AI models," says Oracle founder, executive chairman, and CTO Larry Ellison during the call. "Soon, Oracle will begin construction of data centers that are more than a gigawatt." With this vast number of data centers, one of Oracle's most significant expenses is likely electricity. With global warming and the carbon footprint on many people's minds, the company needs to find economical sources of green energy -- something that an SMR could potentially deliver. Oracle isn't the first company to look into nuclear power, as Microsoft has reportedly begun searching for a critical figurehead for its nuclear strategy. Nevertheless, the former has taken the next step in acquiring building permits. But don't expect to see tiny nuclear reactors popping up around the country next year just yet. After all, it will take years before a nuclear power plant, even a tiny one, can go from a hole in the ground into a fully functional electrical source. The Register even said that the most optimistic estimates for when SMRs will begin deployment are in the early 2030s -- some seven to ten years from now. In the meantime, data centers would have to find other energy sources to power their AI dreams.

Oracle to power 1GW datacenter with trio of nuclear reactors

Oracle is going nuclear over growing demand for AI datacenters, and that's not a metaphor for Larry Ellison's mood. On Monday's quarterly earnings call, Oracle's founder, chair and CTO revealed the database giant and cloud provider had secured building permits for a trio of small modular reactors (SMRs) to power a datacenter with over a gigawatt of AI compute capacity. If you're not familiar, SMRs are miniaturized nuclear reactors not unlike those used in US naval vessels for nearly 70 years. Unlike marine reactors, SMRs are designed to be mass produced. And unlike conventional reactor designs they won't require massive physical infrastructure. In theory they'll therefore be less expensive to operate, but still capable of producing tens or hundreds of megawatts of energy. But no SMR is currently operating and pilot projects have not gone well. Oracle's interest in SMRs as a power source comes as the cloud provider looks to expand its datacenter footprint. "Oracle has 162 cloud datacenters, live and under construction throughout the world. The largest of these datacenters is 800 megawatts, and it will contain acres of Nvidia GPU clusters able to train the world's largest AI models," Ellison told analysts on the call. "Soon Oracle will begin construction of datacenters that are more than a gigawatt." Execs did not say when the gigawatt-class datacenters and the SMRs powering them will come online. However, we've previously looked at SMRs as a datacenter power source, and even the most optimistic estimates put the first deployments in the early 2030s. We've asked Oracle for comment and will let you know if the mega-corp responds. Despite their potential benefits, SMRs face considerable barriers to widespread deployment. Back in May, the Institute for Energy Economics and Financial Analysis concluded that SMRs were "too expensive, too slow to build, and too risky to play a significant role in transitioning away from fossil fuels." However that hasn't stopped major hyperscalers from embracing nuclear reactors - whether SMRs or conventional designs. Earlier this year Amazon acquired Talen Energy's Cumulus datacenter in a $650 million deal. Co-located alongside the 2.5 gigawatt Susquehanna nuclear power plant, the acquisition guarantees the cloud giant access to as much as 960 megawatts of capacity. Meanwhile, Microsoft is also sufficiently interested in SMRs that it hired someone to oversee their deployment. ®

Oracle To Deploy A Supercluster Of ~130,000 NVIDIA Blackwell GPUs, Alludes To A "Gigawatt" Capacity Data Center That Will Be Powered By 3 Nuclear Reactors

This is not investment advice. The author has no position in any of the stocks mentioned. Wccftech.com has a disclosure and ethics policy. Oracle's latest earnings call has served as an eye-opener of sorts for AI skeptics, with the company taking pains to assert that the demand for data centers is still "outstripping supply." What's more, some of Oracle's upcoming facilities are truly mind-boggling, employing NVIDIA's GPUs at a scale that has never been seen before. For the quarter that ended on the 31st of August, Oracle reported solid results a few hours back, managing to beat consensus expectations on both its top-line and bottom-line metrics. What's more the company's AI-critical cloud division recorded an year-over-year growth of 21 percent. During the accompanying earnings call, Oracle wowed investors by noting that the demand for its products and services, particularly in the cloud sphere, was "still outstripping supply" and that the company was now trying to overcome this deficit by "laying out a lot of supply." Apart from building a lot of data centers, Oracle is also rolling out services such as GenAI Agents, which combine Large Language Models (LLMs) with retrieval-augmented generation technology, allowing clients to directly interact with their cloud-stored data in new and innovative ways. According to Oracle, building a foundational model such as ChatGPT or Google's Gemini from scratch will now cost $100 billion "over the next 4, 5 years." Therefore, Oracle believes that the next big growth in AI-related spending will come from the inference process: "So that goes on, and we'll see more and more applications look at that. So I wouldn't -- if your horizon is over the next 5 years, maybe even the next 10 years, I wouldn't worry about, hey, we've now trained all the models we need and all we need to do is inferencing. I think this is an ongoing battle for technical supremacy that will be fought by a handful of companies and maybe one nation state over the next 5 years at least, but probably more like 10. So this business is just growing larger and larger and larger. There's no slowdown or shift coming." This growing demand paradigm is, of course, bullish for Oracle's medium-term prospects. Also, Oracle has just announced a "zettascale" supercluster that "can scale up to 131,072 Blackwell GPUs with NVIDIA ConnectX-7 NICs for RoCEv2 or NVIDIA Quantum-2 InfiniBand networking to deliver an astounding 2.4 zettaflops of peak AI compute to the cloud." This new supercluster is expected to come online by next year and will help clients train and deploy next-gen AI models. Finally, do note that Oracle is planning to build as many as 2,000 data centers, constituting a phenomenal increase from its current capacity of 160! Also, one such facility will consume at least a "gigawatt" of power and will be powered by as many as three small-scale nuclear reactors (SMRs).

Oracle is drawing up plans to power a data center with nuclear reactors

Computing giant Oracle (ORCL) unveiled quarterly earnings on Monday. Revenue was $13 billion, up 8% from a year ago; net income was $2.9 billion, up 23%. Though shares rose more than 11% after the report during Tuesday trading, the occasion revealed another eye-catching development from its founder Larry Ellison. "We're in the middle of designing a data center that's north of a gigawatt," he said on the company's earnings call. "The location and the power place we've located, they've already got building permits for three nuclear reactors. These are the small modular nuclear reactors to power the data center. This is how crazy it's getting. This is what's going on." Artificial intelligence training models and the giant computer warehouses that run them and the rest of the internet are very expensive, electricity-wise. Arm Holdings (ARM) chief marketing officer Ami Badani said in April that data centers eat up 2% of all global energy needs. That number is almost certain to grow if the technology continues to become more widely adopted. "We won't be able to continue the advancements of AI without addressing power," Badani said. "ChatGPT requires 15 times more energy than a traditional web search." In a report earlier this year on expected electricity demand, the International Energy Association said that data centers will make up a third of new energy needs in the U.S. through 2026 and more than double worldwide by 2026 to 1,000 terawatt-hours. "This demand is roughly equivalent to the electricity consumption of Japan," the IEA said. "Updated regulations and technological improvements, including on efficiency, will be crucial to moderate the surge in energy consumption." Oracle isn't the only tech name getting into the nuclear energy game. Amazon Web Services (AMZN) explored a deal to use some of it to power its data centers, and Microsoft (MSFT) co-founder Bill Gates has invested in the nuclear power firm TerraPower to help create more carbon-free energy generation.

Oracle is designing a data center that would be powered by three small nuclear reactors

A view of Oracle's headquarters in Redwood Shores, California, on Sept. 11, 2023. Oracle chairman and co-founder Larry Ellison had a "bizarre" announcement to make this week. The electricity demand from artificial intelligence is becoming so "crazy" that Oracle is looking to secure power from next-generation nuclear technology, Ellison told investors on the company's earnings call Monday. "Let me say something that's going to sound really bizarre," Ellison told analysts. "Well, you'd probably say, well, he says bizarre things all the time, so why is he announcing this one. It must be really bizarre." Oracle is designing a data center that will require more than a gigawatt of electricity, the company's chairman said. The data center would be powered by three small nuclear reactors, he added. "The location and the power place we've located, they've already got building permits for three nuclear reactors," Ellison said. "These are the small modular nuclear reactors to power the data center. This is how crazy it's getting. This is what's going on." Ellison did not disclose the location of the data center or the future reactors. CNBC reached out to Oracle for comment. Small modular nuclear reactors are new designs that promise to speed the deployment of reliable, carbon-free energy as power demand rises from data centers, manufacturing, and the broader electrification of the economy. Generally, these reactors are 300 megawatts or less, about a third the size of the typical reactor in the current U.S. fleet. They would be prefabricated in several pieces and then assembled on the site, reducing the capital costs that stymie larger plants. Right now, small modular reactors are a technology of the future, with executives in the nuclear industry generally agreeing that they won't be commercialized in the U.S. until the 2030s. There are currently three operational small modular reactors in the world, according to the Nuclear Energy Agency. Two are in China and Russia, the central geopolitical adversaries of the U.S. A test reactor is also operational in Japan.